Proof that Alexa rankings can be very inaccurate for small websites

Here is a comparison of the actual viewing figures for this site (as recorded in Google Analytics), against the values that Alexa reports.

Alexa vs Google Analytics page views

This is a fairly small blog, getting 600-800 page views per day. If you too have a website which is of a similar size you might be interested in just how wrong Alexa’s statistics can be. Here is Alexa’s estimate as to how the number of page views I’m getting has changed in the last three months (snapshots all taken on the 15th July).

Looks pretty amazing right? A 2420% increase in page views in three months. But all is not as it seems. Here are my actual page views from Google Analytics (same three month period). You can see the actual figures show a very gentle increase in page views.

Alexa operates by sampling data from the people who have installed its toolbar, or who have installed (or had installed for them!) some other piece of software which sends data back to Alexa.

The percentage of internet users who have Alexa software on their computers is small so the data is subject to anomalies, especially as the website gets smaller. A large website such as Facebook or MySpace probably gets ok results from Alexa as there will be thousands of people browsing these sitse who have the toolbar installed. But the smaller the site, the bigger the potential for the results reported by Alexa to be out of sync with the actual results.

Alexa vs Google Analytics page views per person

So why the huge reality vs Alexa distortion for my site? I’ll come to that later. Next let’s looks at another graph. This is how Alexa thinks my page views per person is doing.

Pretty impressive! The page views per person has gone from 1-2 all the way up to 30 pages viewed per user. If that were accurate it would mean that people really love reading what I write! But unfortunately my blog isn’t quite that engaging. Here are the real figures. Each of my visitors looks at about 1.3 pages. That figure has remained fairly flat over the last three months.

Alexa vs Google Analytics visitors per country

Here is my final set of figures. This is where Alexa thinks most of my visitors are coming from.

You can see from their figures that they think most (74%) of my visitors are in Japan. It is vaguely plausible as I do have quite a few Japan related postings on this site. However it isn’t accurate. Here are the real figures.

You can see that only 1.4% of my visitors are in Japan. This largest percentage of viewers is actually from the UK (44%), but Alexa only believes that 4.5% of my viewers come from the UK.

So why is Alexa so badly wrong?

What is causing Alexa to be so badly out of sync from reality? I am pretty sure I’ve worked out the reason. I believe that the massive inaccuracies are the result of a single person. Me.

Not on purpose, I did it entirely accidentally. But it is interesting that one person could cause the perceived page views of a 600-800 page per day blog to jump 2420% in three months. It is also interesting that one single person could accidentally cause the Alexa traffic rank to jump from about 600,000 to 200,000 in such a short time. And it is impressive that one single person could cause Alexa to think that 74% of visitors are from Japan when in fact only 1.4% of them are. Also that I could cause the average page views per person to go from 1-2, all the way up to 30ish.

How did I (accidentally) cause Alexa’s figures to be so wrong?

There is one main cause of this massive misreporting of figures. I installed the Quirk SearchStatus extension for Firefox. This extension add some small items to your status bar at the bottom of the browser. They can show the Google PageRank for the web page you are currently on, and they can show the Alexa traffic rank for the website you are on.

But the extension doesn’t silently report the statistics. It actually affects the Alexa traffic rank statistics. As you are browsing where you go will now be recorded by Alexa.

In my case I spend a lot of time on my own blog. It is WordPress blog so I need to access it to put up new posts, edit posts, check posts, moderate comments, and install/configure new/updated plugins. In a typical day this could easily generate another 30-50 page views of my blog that are reported to Alexa.

It would seem that although all my page views are coming from the same IP address, they are still able to make Alexa think that my blog is seeing a huge surge in traffic.

The timing is right too. The time I installed the plugin matches the time I started seeing this huge surge in Alexa rankings. And that’s not all. A short while after installing the plugin I headed to Takamatsu in Japan and continued working on my blog from there.

The fact that I had moved from the UK to Japan caused caused Alexa to think that most of my visitors were now coming from Japan.

Almost all of this was caused by me accessing my WordPress blog in a normal way whilst having this Firefox extension installed.

What it shows is how inaccurate the numbers can be for small sites such as mine. The number of people whose browsers report statistics to Alexa is sufficiently small that a single person can seriously skew the data.

So look at Alexa’s statistics if you really want to. But remember that unless you are looking at the stats for a huge site, they may well be way off the mark.